基于BP神经网络的水轮机调节系统非线性模型及仿真计算

Modeling and Simulation of Water Turbine Nonlinear Regulation System Based on BP Neural Network

针对水轮机调节系统仿真中采用线性水轮机模型时仿真结果与实际运行有一定偏差或采用基于插值非线性模型时插值难以保证处处连续并导致压力环迭代不收敛的问题,先利用BP神经网络对水轮机运行数据进行延拓和训练,以获取连续光滑的流量和力矩空间曲面供压力迭代计算使用,再结合特征线算法合理选取边界条件,最终建立起完整的调节系统高精度非线性模型。仿真结果表明,该模型具有相当高的精度,仿真效果与预期一致。

To solve the problem of the general linear water turbine model that widely applied in the hydro-plant simulation, but the simulation results differ with the actual situation. What＇s more, the continuity of non-linear model based interpolation can hardly be controlled that could lead to the non-convergence of the pressure loop iteration. This paper firstly adopted the BP neural network to get the continuous and smooth space surfaces by training the operation data of the specific water turbine and applied them to the pressure iteration and interpolation. And then it established the nonlinear model of the whole regulation system with proper boundary conditions by combining with the characteristic method. The simulation results are consistent with our expectations, which verifies the high accuracy of the model.